应用神经网络和AE信号对磨削烧伤的在线检测

Grinding Burn Online Detected by Neural Network and Acoustic Emission

磨削烧伤是磨削过程中常见缺陷之一,严重影响被加工零件质量和使用寿命,运用RBF神经网络和AE传感器实现了磨削过程中磨削烧伤的在线检测,通过分析磨削加工中AE信号的特性,计算240～400kHz内的信号有效值,峭度和歪度,处理后作为神经网络的输入向量,完成磨削烧伤的在线识别,通过比较在线识别结果和离线检测结果,证明了该在线检测系统具有较高的准确性.

Grinding burn forms frequently in grinding process, it decreases quality and the useful life of workpieces. A method is proposed to detect the workpiece burn online in grinding process by RBF neural network. The grinding acoustic emission （AE） signals were collected and digested to extract feature vectors that appear to he suitable for neural network processing. The feature vectors, which consists of band power, kurtosis and skew were the statistics extracted from the 240 kHz to 400 kHz AE Signals. Compared the results of offline testing with the results of online detecting, this online detecting system was proved efficient and accurate.